The Pywikibot tests are based on the unittest framework.
The tests package provides a function load_tests that supports the load tests protocol. The default ordering begins with tests of underlying components, then tests site and page semantics, and finishes with tests of the scripts and finally any tests which have not been inserted into the ordered list of tests.
A function collector also exists in the 'tests' package.
The entire suite of tests may be run in the following ways from the root directory:
pip install pytest-runner python setup.py pytest
python -m unittest discover -v -p "*_tests.py"
py.test
tox
Individual test components can be run using unittest, pytest or pwb. With -lang and -family or -site options pwb can be used to specify a site.
python -m unittest -v tests.api_tests tests.site_tests python -m unittest -v tests.api_tests.TestParamInfo.test_init
py.test -s -v tests/api_tests.py tests/site_tests.py py.test -s -v tests/api_tests.py::TestParamInfo::test_init
python pwb.py tests/api_tests -v python pwb.py tests/site_tests -v python pwb.py tests/api_tests -v TestParamInfo.test_init python pwb.py -lang:de -family:wikipedia tests/page_tests -v TestPageObject
PYWIKIBOT_TEST_MODULES=api,site python -m unittest -v
After changes are published into a GitHub repository, tests may be run on a Microsoft Windows box provided by ci.appveyor.com according to the configuration in .appveyor.yml file. To do this:
- create a GitHub and AppVeyor account
- fork the main GitHub repository
- create a project in ci.appveyor.com
- go to https://ci.appveyor.com/project/<username>/pywikibot/settings and enter the custom configuration .yml filename: .appveyor.yml
- push changes into the forked git repository
- watch the build at https://ci.appveyor.com/<username>/pywikibot/history
The 'user' tests are not yet enabled on AppVeyor builds.
There are a set of 'edit failure' tests, which attempt to write to the wikis and should fail. If there is a bug in pywikibot or MediaWiki, these tests may actually perform a write operation.
These 'edit failure' tests are disabled by default. On Travis they are enabled by default on builds by any other GitHub account except 'wikimedia'.
To disable 'edit failure' tests, set PYWIKIBOT_TEST_WRITE_FAIL=0
There are also several other 'write' tests which also attempt to perform write operations successfully. These will write to the wikis, and they should always only write to 'test' wikis.
These 'write' tests are disabled by default, and currently cannot be run on Travis or AppVeyor as they require interaction using a terminal. Also enabling them won't enable 'edit failure' tests.
To enable 'write' tests, set PYWIKIBOT_TEST_WRITE=1
Enabling only 'edit failure' tests or 'write' tests won't enable the other tests automatically.
pywikibot's test suite, including Python's unittest module, provides decorators to modify the behaviour of the test cases.
Skip a test if the condition is true. Refer to unittest's documentation.
import unittest [......] @unittest.skipIf(check_if_fatal(), 'Something is not okay.') def test_skipIf(self):
Skip a test unless the condition is true. Refer to unittest's documentation.
import unittest [......] @unittest.skipUnless(check_if_true(), 'Something must happen.') def test_skipUnless(self):
Require that the given list of modules can be imported.
from tests.aspects import require_modules [......] @require_modules(['important1', 'musthave2']) def test_require_modules(self):
Replaces target with object specified in new. Refer to mock's documentation. This is especially useful in tests, where requests to third-parties should be avoided.
from tests import patch def fake_ping(url): return 'pong' [......] @patch('http_ping', side_effect=fake_ping) def test_patch(self): self.assertEqual('pong', http_ping())
Test modules should be named according to the pywikibot that is being tested. e.g. the module pywikibot.page is tested by tests.page_tests.
New test classes should be added to the existing test modules unless it tests a new component of pywikibot.
All test classes must be a subclass of tests.aspects.TestCase, which uses a metaclass to dynamically check the test can be run on a specified site, or run a test on multiple sites.
If a test depends on a specific site, add class attributes 'family' and code'.
family = 'wikipedia' code = 'en'
Once declared, the Site object can be accessed at self.site.
If a test requires multiple specific sites, add a class attribute 'sites'.
sites = { 'enwiki': { 'family': 'wikipedia', 'code': 'en', }, 'itwikt': { 'family': 'wiktionary', 'code': 'it', } }
To obtain the Site object, call self.get_site with the key given to the site.
self.get_site('itwikt')
For tests which require network access to a website which is not an APISite, the class attribute 'sites' may include a hostname.
sites = { 'wdq': 'hostname': 'wdq.wmflabs.org', } }
net = False
: test class does not use a sitedry = True
: test class can use a fake site objectcached = True
: test class may aggressively cache API responseslogin = True
: test class needs to login to sitesysop = True
: test class needs to login to site as a sysopwrite = True
: test class needs to write to a site